Questionable puzzle

I think this puzzle has two solutions: lichess.org/training/QJH5D (Qe8 and Qxd4)

It asks you to play "the best move". The best move on the board is definitely Qe8

Qe8 is a mate-in-one.

Qd4 goes into an ending with connected passed pawns vs bishop.

I think they are not even remotely close in evaluation.

I can see that one of the moves is mate, but is there some reason to allow puzzles with multiple solutions?

Simply asking for the "best move" has the following drawbacks:
1, one would need to specify criteria by which the best move should be determined. Which in this case would be the name and version of the engine, analysis time, hardware specs etc.
2, it would essentially turn every position into a puzzle by simple asking to find the top line of some engine, ran for a specific time on specific hardware.

I understand that puzzles are generated algorithmically, which is why human quality control is also important.
Even so, it shouldn't be too hard to check that a position doesn't have multiple winning/drawing moves.
Most lichess puzzles follow this policy which is why it seemed like this one slipped through the cracks...

There are cases, occasionally, when the difference between the best move/line and second best is rather small and when that happens, the puzzle is removed. But here the difference is huge. If it was a real game and you had to choose between an immediate checkmate and an endgame where you are winning but you would still need a lot of effort to convince your opponent about that (while having to be careful not to lose due to one blunder), would you really consider choosing the latter?

I agree with OwenKraweki.

There was a similar case flagged in this forum a couple of weeks ago here: lichess.org/forum/lichess-feedback/i-think-puzzle-uutff-should-be-removed

In that case I thought it was borderline because the alternative move (from a position a minor piece down winning back a minor piece instead of queen for a second minor piece) was possibly not quite clearly winning.

But in this case it looks to me at first sight that 38.Qxd4 is an absolutely clear alternative win.

Perhaps the requirement that a puzzle should have only one winning move is relaxed, then, if one move stands out as far "better" in the sense of much more quickly winning? I'm guessing now that the puzzle generator must have something like that built into it.

EDIT: cross-posted with mkubecek.

There are certainly puzzles with more than one winning move, e.g. when everything except stupid blunders still leaves you with a solid position and two pawns up but there is one clever tactics ending with a checkmate or queen/rook up. And I believe this is perfectly fine. What I personally don't like much are puzzles where even the best move gives you only a small edge like a pawn or two or more active position. With these, I'm always unsure if what I've found is really the solution or if I missed something better. (Especially if there are multiple checkmate sidelines but in the main one the king escapes at the expense of a minor material loss.) But I guess even that makes sense to make the experience more lifelike.

It would even make sense to have puzzles where the best move is not winning at all, just saving a draw in seemingly losing position. But I'm not sure if I have seen such puzzle here yet.

@mkubecek said in #7:
> There are certainly puzzles with more than one winning move, e.g. when everything except stupid blunders still leaves you with a solid position and two pawns up but there is one clever tactics ending with a checkmate or queen/rook up. And I believe this is perfectly fine. [...]

Yes, so it seems. I agree it's fine. I think they must be quite rare though because I don't seem to encounter them.

> It would even make sense to have puzzles where the best move is not winning at all, just saving a draw in seemingly losing position. But I'm not sure if I have seen such puzzle here yet.

Yes, these would be an important addition to the site. The fact that all puzzles seem to be "find a win" rather than a draw is probably the reason why "stalemate" or "perpetual check" are not on offer as puzzle themes.

I can only say again I quite strongly disagree in this case. A mate in one ends the game right there.

The resulting position after Qxd4 will give two connected passers on the queenside, and on the kingside either a 2-vs-1 majority with the possibility of creating another passed pawn, or a crippled white pawn structure.

There will be a lot of work to do to win this position, and the fight will go on for quite some time, and there are serious chances for a draw if you play inaccurately.

I have said it before in another thread: I really like that you don't always have to find the only winning move, but the best move. This makes it much more relevant for practical play. This game from the puzzle is a good reason why. In a 3+0 game with around 25 seconds on the clock, the practical chance (or risk) of not winning after Qxd4 are very real.

Yeah I can see the arguments for both sides. Sure it doesn't end right away, but Qxd4 is a clear alternative win for me too. Qe8 is also the best move undoubtedly.

The initial eval for Qxd4 is not so high however. I didn't check with higher depths. But algorithmically probably this is why this puzzle was not discarded. If it were to be evaluated as +5, I doubt this puzzle would make it even after having mate in 1.